Microsoft's role in Sam Altman's return is no longer just a Silicon Valley memory. It is now evidence in a trial about who really controls the most important AI startup in the world.
Satya Nadella's appearance in federal court on May 11 put a sharper frame around Elon Musk's case against OpenAI: the argument is not only about a nonprofit that became a commercial powerhouse, but about how quickly a strategic investor can become the gravity point around a founder-led company.
Musk's lawyers are trying to show that OpenAI, Sam Altman, Greg Brockman and Microsoft moved the company away from its original nonprofit mission and toward a structure that enriched insiders and gave Microsoft unusual influence. Nadella's answer was simple. Microsoft backed OpenAI heavily, but did not control it. That distinction now sits at the center of the case.
According to the Wall Street Journal's coverage of the trial, Nadella testified about the frantic days after OpenAI's board fired Altman in November 2023, saying Microsoft was worried about losing key AI talent and protecting a multibillion-dollar investment. That is not a small fact. When an investor has committed roughly $13 billion, has built products around the startup's technology and can offer jobs to the team overnight, formal control is only one part of the power equation.
The courtroom question is whether Microsoft crossed a legal line. The business question is more useful for founders. How much influence does a partner need before board independence becomes theoretical?
Nadella reportedly denied that Microsoft controlled OpenAI and said Musk had opportunities to object to the Microsoft relationship earlier. OpenAI has also argued that Musk is now a competitor through xAI and that his case is less about mission purity than business rivalry. Those points matter, because Musk is not a neutral observer. He helped found OpenAI, later left, and now runs one of its most visible challengers.
Still, the 2023 firing exposed a weakness that every startup board should study. OpenAI's nonprofit board had the legal authority to remove Altman. It used that authority. Within days, employee pressure, investor pressure and Microsoft's willingness to absorb talent helped reverse the decision. Altman returned. Several directors left. The board had power on paper, but not enough institutional support to make its decision stick.
That is the kind of governance failure founders often ignore while capital is easy to raise. Everyone is aligned until the first serious conflict. Then contracts, voting thresholds, information rights and employment leverage become the company.
OpenAI's later restructuring adds another layer. In October 2025, OpenAI said it had completed a recapitalization that left the nonprofit, now called the OpenAI Foundation, in control of OpenAI Group PBC, a public benefit corporation. The company said the foundation's equity was valued at about $130 billion and that the mission of ensuring artificial general intelligence benefits humanity would continue through both the business and the foundation.
That sounds elegant. It may also be harder to test in practice.
Harder to fire can mean harder to govern
New court reporting has focused on a bylaw change adopted during the 2025 restructuring. Business Insider reported that Musk's expert witness pointed to rules requiring a two-thirds vote of nonemployee directors of the public benefit company to remove the CEO, compared with the simple majority standard that applied under the prior nonprofit board structure.
There is a reasonable defense of that kind of rule. AI labs are not ordinary software companies. A sudden CEO removal can trigger staff exits, customer panic, partner uncertainty and government scrutiny. Stability has value, especially when the company is trying to raise enormous sums for compute, chips, data centers and research talent.
But there is an obvious tradeoff. A supermajority removal rule can protect a company from a reckless board. It can also protect management from a responsible board. The same mechanism that prevents chaos can create entrenchment if directors lack access to information, courage or independence.
That is why OpenAI's structure matters beyond OpenAI. The next generation of AI startups will not grow through clean venture rounds alone. They will need cloud credits, chip access, distribution deals, model partnerships and, in some cases, regulatory confidence. The partner that brings those resources may never need majority ownership to shape outcomes.
Founders should take the lesson early. Board design is not paperwork to finish after the term sheet. It is the operating system for the worst day of the company. If a strategic investor becomes essential to infrastructure, revenue and hiring, the governance documents must answer what happens when that investor's needs diverge from the mission.
OpenAI's trial is full of personalities, and that makes it easy to watch as courtroom drama. But the lasting issue is more practical. The AI companies that survive will need capital and discipline at the same time. If their boards cannot act when the mission and the money pull in different directions, the real control will sit with whoever the company cannot afford to lose.
Also read: Amazon's AI push shows how bad metrics can distort good tools • Artificial Analysis shows coding agents are more than model scores • Unitree turns a mecha demo into a robotics startup test
